I’m posting this here instead of on X this period, to keep the discussion more internal to the involved community while we are cooking something new.
Intro and summary:
I have personally been AI-pilled since 2010, when I did a visiting fellowship at the Machine Intelligence Research Institute. As much as AI feels like a buzzword at this point, and as silly as some of the valuations we’ve seen may seem, I actually think most people still don’t understand how much it could really revolutionize everything we experience. To me, the scenario of centuries of scientific progress in a decade is not that far-fetched. But nor do we, as a society, appreciate how disruptive and destructive to our current way of life it could be, depending on how the technology is built out and wielded. I really don’t want any single person, company, country, or agent to achieve a decisive scientific and technological advantage over everyone else that could give it a perpetual monopoly on violence and position of absolute coercive control.
Those who share my views about the power of AI were early investors in AI startups. More recently, they’ve been early to the AI supply chain trade. Fortunes have already been made.
And now, because many of the companies building out the AI revolution are public, we’re starting to see mainstream retail participate in this investable moment. The DRAM ETF is the starkest example so far. It offers the top memory producers in the AI supply chain, including two South Korean companies that U.S. retail had a hard time accessing before its release. In a little over a month, it went from zero to about $9 billion in AUM, with virtually no marketing other than the organic coverage it’s received – one of the fastest-growing ETFs of all time.
U.S. equity tokenizers have taken notice, and this category is about to come on chain. Reserve is going to lead the way in offering DTF exposure to the category of AI and the subcategories of the AI supply chain. While blockchain-based AI projects may have a place as this all progresses, what’s more certain is that real companies building and supplying this rollout are doing very well.
ABC Labs is going to launch a series of geo-restricted AI thematic equity DTFs this summer, built on tokenized US stocks. Photonics first, perhaps power second, more after that, with an option to also offer a generic AI infrastructure DTF that captures the entire supply chain.
About 75% of the team is shifting focus to the first launch over the next four weeks while our research pod continues developing the parallel crypto DTF product line.
We posted the milestone-based RSR unlocking proposal on the forum for community discussion (any unlock gated on hitting $2.5M in net annualized recurring revenue). Engineering shipped the foundations for optimistic governance integration on the Reserve front end and continued building out the agent-managed DTF MVP, the new data indexer, and Reserve SDK v0.1. There is a lot in this update; the AI DTF launch is the headline and is worth reading the most carefully.
ABC Labs is going to launch AI thematic equity DTFs
The idea has been brewing for a few periods. At our onsite, we discussed both (a) how to leverage AI and tokenization for asset-backed currency, and (b) whether we could offer real exposure to the AI industry onchain. It resurfaced at the start of this period with an “AI Exposure Fund” exploration at the ABC Leads Weekly two weeks ago, deepened through two AI Fund coworks the following days, and crystallized in calls with tokenized-stock issuers earlier this week. I walked the full team through the direction at the Reserve weekly sync this past week, and I’m excited to share it with you here.
What we are building: Two DTF product lines this summer.
The first is a set of AI thematic equity DTFs, built on tokenized US stocks. One DTF per AI sub-category. Photonics first, perhaps power second, more after that, and perhaps a generic AI or AI infrastructure DTF that captures the whole category. These will all be geo-restricted to mirror whatever the underlying tokenizers require, which in practice means non-US, with APAC and LATAM as the near-term retail markets, and potentially accredited investor access in the EU and some other regions. We plan to launch the first one in June, on the same day that the underlying stocks are released in tokenized form.
The second is a set of crypto DTFs. BTC/Gold and others. These will be worldwide. The timeline is over the summer while we are still in the bearish part of the cycle, because the early life of these DTFs is much more interesting if users get to ride a recovery from a low entry point.
About 75 percent of the team is shifting focus to the AI launch for the next four weeks. The research and fundamentals group will keep working on the crypto line in parallel. I do not want to disturb their work, because I think both product lines should ship roughly together this summer, and we can do the AI launch without distracting them.
Why AI, and why now:
Three reasons: immediate demand, long-term value, and feasibility just clicking into place.
On demand: the AI thematic trade in public equities has moved past NVIDIA and the model companies, and the money is flowing down the supply chain. Memory (SanDisk, SK Hynix, Samsung). Photonics (which Goldman Sachs put a research spotlight on with their April 17 Optical Networking report). Power infrastructure (GE Vernova, Wolfspeed, the whole turbine and fuel-cell story). And now a long tail of obscure component suppliers. The DRAM ETF launched a month ago and now has $9.69B in AUM. Situational Awareness and Value Aligned Research (which Confusion Capital invested a small portion of our treasury in last year) are each sitting on multiple billions in AUM playing variants of this trade. There is real money invested in this category.
The trade has also gone public on X. Accounts like Serenity, Citrini, and Primer publish supply chain analysis that used to live inside hedge funds with $30K Bloomberg terminals. Ivan Camps (Head of GTM) has been tracking the social side closely. Eric Fang (Treasury Manager) is replicating a meaningful chunk of that workflow with Claude in his free time, feeding in earnings transcripts, equity research, and supply chain pieces and using them to build a working picture of where the next bottleneck is. He demoed it to the team at the weekly sync and shared a Claude artifact visualizing the value chain.
You can see the public interest in this trade by the virality of the posts and the meaningful appreciation of assets in each new category of the AI supply chain that this community starts talking about. Many of the companies involved are from the U.S., so for U.S. people, you can just open your brokerage account and buy whichever part of the infrastructure stack you think is due to appreciate.
Investors abroad may not have it so easy. Some countries have regulations in infrastructure that make U.S. markets fully accessible, but others don’t. This is the segment of investors that equity tokenization companies have targeted with on-chain versions of U.S. stocks, and the growth in assets shows that it’s working.
Our bet is that there are hundreds of millions, if not billions, of dollars in demand for access to this category of U.S. equity exposure abroad that are running into trouble finding a way to purchase. A meaningful portion of that capital is held by people and funds that are either crypto-native or crypto-comfortable. Think about it: why do most people get into crypto investing? It’s because they’re comfortable with volatility and they’re looking for market-beating returns. Right now, AI stocks are more exciting to that population than the crypto category.
We’re betting that there are meaningful pockets of pent-up demand for these assets among purchasers who will be happy to connect their wallet and hold them in self-custody. We’re in touch with the tokenizers of these equities and believe that at least one of them is on track to serve this category of assets in the near term, making it possible for the first time to construct on-chain index products for each category of the AI supply chain trade.
Is this 1999?
My take: yes and no.
The early Internet was a true revolution, but it took a decade to integrate into how we live. Part of that is because the substrate (broadband, cloud, mobile, smartphones) had to be built before the businesses on top could fully work. AI is different. It’s a multiplier on top of the current stack. When a new AI capability lands, it gets integrated into existing software in months. That compresses the gap between speculation and real revenue.
That does not mean it is a straight line up. I expect huge volatility. Companies that have 10xd could lose 95 percent of their value as AI paradigms shift. Maybe we won’t need all of the electricity we think we do. The supposed sub-quadratic attention breakthrough is a good example: if a small lab can match Transformer performance at roughly one twentieth of the cost by having the network strategically attend to fewer tokens at a time, the entire power infrastructure thesis could get pulled out from under us. And that’s a relatively small architecture change – there may be much bigger breakthroughs in the next few years. The cycles within the category may also be faster than what we are used to in crypto, because narratives are tied to actual earnings calls and product releases, not just self-referential speculative momentum, which has broadly swung back and forth over a four-year period.
I think the category compounds over the next decade. And in the meantime, I think the speculation around it is going to look a lot like crypto speculation in 2017 or 2021. Big swings, lots of attention, communities forming around specific names. It will no doubt be hard to keep up with all of the madness and tell whether any particular company is overpriced or underpriced.
Of course nobody knows for sure how this will play out. We could be disappointed by the lack of breakthroughs five years from now, we could be enjoying the automation of all mundane tasks, or we could be staring into the singularity.
Why an index, not active management:
Imagine you have a strong view that photonics is the next big bottleneck. Even with that view, picking individual names is tough. Some will 10x, some could go down 90 percent, and you have to hold through all the noise. An index over the whole category gives you the thesis exposure without the single-name risk.
Actively managed funds could make sense, but would require more prep time, talent acquisition, and regulatory work. For now, the task of identifying and defining the categories and methodologies will keep us busy. And it would appear there are plenty of gains to be had from this disciplined approach.
This is different from the methodology we have been building for the crypto DTF line. With BTC/Gold and the other strategies we have been researching, the question is whether a statistical pattern from history will hold up out of sample. With the AI thematic DTFs, the question is whether the underlying companies are going to be more valuable in the future than they are today as this industry builds itself out. That is a thesis question, not a backtest question. The capacity we have been building up for backtesting and statistical robustness does not really apply here, which is what frees us to run this launch in four weeks while keeping the research pod focused on the crypto line.
Who this is for:
The starting audience is non-US, crypto-comfortable investors who want exposure to AI supply chain names and cannot easily get them through the legacy stock market. APAC and LATAM are the near-term focus.
The crypto-native angle here matters. We’re betting there’s a huge crossover between people who are comfortable transacting on chain because they have a history of investing in crypto, and people who want to invest in this category and participate in this level of volatility.
The week-1 strategy and the chicken-and-egg problem: This is the part I have been thinking about the most. If we hit $100M in week 1, everyone talks about it and we run to $1B by week 3. If we hover at $5M, few will talk about it and we’ll still be at $5M at week 3. The slow path works fine over 6 to 12 months, but a massive week-1 breakout could unlock outsized growth essentially for free. So we are spending real prep time on how to solve the chicken-and-egg within the first three days of being live: getting the most-followed AI supply chain analysts to talk about it, sizing seed TVL from our own treasury, working on external seed capital, and being willing to spend on paid amplification in weeks 1 and 2 if the plan looks promising. This is the kind of moment I was saving the budget for by winding down incentive spend on the other DTF lines. The DRAM ETF is the reference case (billions in AUM in its first month with no big marketing push, from a flywheel of free media coverage). We are not promising DRAM-style outcomes; we are studying it carefully.
The problem with these directions is that it’s always easy to think of promo tactics that would be super valuable, but it’s almost never easy to actually land them. It’s easy to say, “If the top AI supply chain research influencers all talked about this on day one, that would be huge.” But why would they be willing to talk about it? It’s easy to say, “If we get a hundred million in week 1, everyone will talk about it and it’ll get way bigger.” But is there any way to reach a hundred million in week 1? I keep waking up a couple of hours before my alarm thinking about this question.
Distribution: Launch channels are app.reserve.org (with geoblocking to match the restrictions of underlying tokens); UGLYCASH (the audience overlap is perfect: non-US, crypto-comfortable, looking for exposure they cannot easily get elsewhere); selective third-party DEX placement where geoblocking can be enforced; and potentially CEX wallet integrations. We are also researching the regulatory feasibility of direct CEX listings in APAC and LATAM.
Chain selection: Relevant tokenized stocks will be on both Ethereum L1 and BNB Chain, and we are leaning BNB. Max Bettinelli (Product) published an ecosystem analysis that landed on the same answer: BNB has more overall TVL with tokenized products and is growing faster than Ethereum for this category, with the long-tail single-name activity driven by direct integrations with Binance and Bitget wallet. Patrick McKelvy (Director of Engineering and Security) confirmed both chains are supportable, so we’re testing on both, but greater than 50 percent odds we go BNB-first.
Geoblocking and compliance: We are blocking sanctioned jurisdictions entirely along with the United States. For a separate set of jurisdictions (Brazil, EU/EEA, Hong Kong, Malaysia, Singapore, Switzerland, UK), we may offer access to accredited or qualified investors under their local definitions. We’re also implementing sanctioned wallet screening.
Liquidity and minting infra: We are building a new backend minting approach that draws on the specialized off-chain sourced liquidity for these tokenized equity products. We’re also planning how to provide liquidity directly on these new DTF tokens in a way that satisfies all compliance requirements. There are several moving pieces in the background that the legal and technical team members are collaborating to tackle on a tight timeline.
Framework: why we think this is a good bet: Internally we evaluate every DTF on four bars: product, brand, advertising, distribution. Roughly 85% confidence on each compounds to about 50% of success overall, which makes it worth spending serious time and budget taking the swing. For the photonics DTF specifically, I think we are already at 85 on product. Brand, advertising, and distribution are the three we have to nail over the next few weeks. The work looks tractable to me, which is why I am willing to make the bet.
Operational plan: Sean Sadasivan (Head of Product) is leading the cross-functional execution with a working AI DTF Launch Plan covering compliance, token ops, governance, branding, marketing, and distribution. Token-ops stress testing is already in flight: Max and Agus have deployed two placeholder mock DTFs and are rebalancing them to prove CoW Swap can handle the tokenized stocks at production-relevant sizes before we go live.
A research product for the AI supply chain investors:
One additional idea on the table: building a website or tool that the population of AI supply chain investors would want to use every day, partly as a marketing surface and partly as something useful in its own right. Two concepts we are looking at: a market tracker that defines 20 indices for the various verticals and narratives so you can track how they are performing relative to one another, and a research platform that aggregates everything Eric is currently scraping with Claude into a Scout-style interface. We do not think this is a go-viral-week-1 option, so for now we are going to have engineering help Eric harden his data scraping setup, and we may productize it later if it earns its place.
Stakes for Reserve:
We have been circling the question for years now of what is the right category, what is the first type of DTF that could get really big and establish the category, what could put a flywheel in motion. I think this category could be the one. AI-related equities have the volatility of memecoins, but they’re based on reality and have real cashflows. Demand is high, and there are people around the world for whom this will be the first way to access the category.
Once we bring in hundreds of thousands of users to one big DTF, the others get easier to pitch. People may come for the AI supply chain exposure and then end up making investments into BTC/Gold because they want a hedge. People who hold two or three DTFs and are comfortable with Reserve might also decide to buy an index of crypto when BTC starts to climb.
This is a very rare setup where you have a lot of attention, a lot of volatility, and potentially a lot of staying power, all in one place. Depending on your view on crypto, the last time this happened was in 1999. The more I have looked at it, the more I have decided we should make a very strong bet here and try to be at the forefront.
[IMAGE: Screenshot from the Reserve weekly sync deck showing the four-bar evaluation framework (product, brand, advertising, distribution) with photonics scored at ~85 percent on product]
RSR unlocking milestone proposal: discussion is open
I posted the milestone-based RSR unlocking plan to the forum last week. The proposal is the one I introduced on the Q1 community call: future RSR unlocks tied to reaching $2.5M in net annualized recurring revenue, where NARR is total recurring revenue minus the incentive spend it takes to generate that revenue. No revenue, no unlocks. The forum thread is here: [RFC] RSR Unlocking Milestone Plan
I will leave it open for at least another couple of weeks before we take next steps, because these things take time to percolate, and because we will be busy with the launch. Eleven voters so far, which is enough for an early vibe check on how people are receiving it.
The most interesting new idea raised in the thread is whether we should track the ratio of net annualized recurring revenue to the total RSR circulating supply, as a way of seeing whether each new RSR coming into circulation is buying us a good amount of revenue. The version I would propose is to track growth in circulating RSR per dollar of growth in NARR. That is conceptually similar to “burn multiple” (which we considered including in the public metrics and decided against because it was too complicated to communicate), but this RSR-denominated version makes a lot of sense in our context. Worked example using the current proposal numbers: we are proposing to add 3 billion RSR to circulation, gated on a $2.5M NARR increase, which works out to about 1,200 RSR unlocked per new dollar of NARR. That seems like a number worth publishing as a comparison metric in future milestone proposals and possibly in quarterly reporting.
Other notes
BTC/Gold launch pulled back: We had been planning to put our BTC/Gold strategy in front of the Top 100 for an internal pitch. Gina ran a much larger walk-forward optimization sweep (180,000 backtests, 9 million combinations), and the result, which Michael walked the team through, was that out-of-sample predictivity collapsed from around 55 percent on a small grid to 6.9 percent on a wider one. That is the search-space-fits-everything problem in action. The research pod (Michael, Nagaking (Senior Data Scientist), Gina Pieters (Lead Research Economist), and Chelsea Canon (Research Manager)) had a long methodology meeting with Sean and I where we worked through the questions in depth, including some real disagreement that the team worked through openly. We decided to pause the internal launch rather than ship a product we could not stand behind with conviction. We will still ship the crypto DTF line over the summer, alongside the AI thematic line. Just not next week.
Team transitions: Griffin Peer (Head of Market Intelligence) transitioned to Securitize this period. Soham Mishra (Institutional Business Development and Partnerships) is in his final week with us before transitioning to Chainlink. We are going to miss them both, and we’ll be in touch as we collaborate with both of those teams in the future. Harold Zavarce is joining as Head of Information & Operational Security, leading info-sec across ABC and CC.
DSI: regulator engagement continues: The Digital Securities Initiative had a productive working session with our outside counsel at Cravath on the SEC transfer agent (TA) relief letter we are preparing for the SEC’s Trading and Markets division. We’ll be presenting to Trading and Markets in the coming weeks. Whereas the tokenized equities we are about to support can’t be touched within the US, we aim to open up regulated securities (from the US and abroad) for use onchain within the US and other strict jurisdictions. The road is still long and uncertain, but a relief letter from the SEC is the next milestone we are aiming for, since it should allow us to convince several teams to embrace an approach together and get down to building in a framework of compatibility rather than everyone doing their own thing.
Engineering ships: The new Reserve SDK (v0.1) is wrapping up, with optimistic governance integration into the Reserve front end as the next step; optimistic governance itself is on track to roll out in Q2. Erik Davis (Senior Designer) tested the first async mints on MCAP, which is a prerequisite for the high-throughput mint flow we will want for the AI launch. The agent-managed DTFs MVP audit prep continues, with Akshat Mittal (Senior Protocol Engineer) driving most of that work. The new data indexer is being integrated into the Zapper, and the rebalance bot is going through an efficiency refactor. AI DTF launch engineering items (sanctioned wallet screening, geoblocking, tokenized stock support in the deployer / pricing API / historical data) are in flight.
Finance ops: Steady work continued on weekly finance reviews, payment forecasting, liquidity and rebalancing across treasury accounts, and crypto bookkeeping (month-end closing). Yuxi Liu (Senior Crypto Accountant), Sasha Kurapova (Staff Accountant), and Kristina Diatchenko (Controller) continue to carry the operational backbone.
Thanks for reading. This is the most excited I have been about a launch in a long time, and it is also the highest stakes we have shipped in a while. The combination of timing, infrastructure availability, and audience overlap with our existing crypto-native user base is rare. We have three to four weeks of focused work ahead and may be hard to reach during that time. However, if you have ideas or suggestions for the launch, you can drop them here and I will check back on this thread periodically.
Macro meta aligned.